How Does Computer Vision Improve Retail?

By Kate Prohorchik, Iflexion

Although it’s not so obvious for the average customer, retail store management is a rather complicated process, which takes the coordinated performance of many different tasks including inventory management, customer support, surveillance, and more. While the day-to-day completion of these tasks is vital for any retail store, they are often boring and repetitive. With the introduction of computer vision, businesses can automate many of such chores, increasing overall efficiency, and freeing up employees’ time. Let’s explore the current applications of computer vision in retail.

Heat Maps

Heat maps allow visualizing the density of activities in a particular area. By installing computer vision-enabled cameras across the store, retailers can better understand what areas the most visited and what products are generate the most interest with their customers. Tracking in-store movement opens up new horizons for companies to better understand their audience and improve store layout as well as product presentation.

Improving The Store Layout

Unless people are familiar with the store and know exactly what products they want to purchase, there is no definitive logic involved in deciding which parts of the shop they are going to visit and in what order. By utilizing heat maps, companies can design store layout based on real data rather than intuition and often inaccurate observations. This way, customer experience can be improved by making the most visited areas more spacious.

Improving Product Presentation

By understanding which exact areas of the store are the most attractive to customers, businesses can make the most profitable products more visible and locate more staff in those areas to improve customer service.

Self-Checkout

With the help of computer vision, the checkout can be done without the involvement of cashiers, significantly reducing workforce costs in the long-term. Although self-checkouts are already widely used by industry giants like Tesco and Walmart, computer vision can take this process to the next level by enabling checkout without barcode scanning.

For example, Abto Software has created a computer-vision-powered tool that can scan up to 130 items in a minute, significantly increasing the checkout rate and reducing average queue times. Contrary to other computer vision projects in retail like Alibaba’s Dragonfly or Amazon Go, Abto Software doesn’t imply huge up-front installation costs and, most importantly, doesn’t compromise customers’ privacy by using facial recognition software, making such solutions much more feasible.

However, any such system is susceptible to the slightest variations in products’ physical appearance. While the AI algorithm can reliably identify every Snicker’s bar, it will need a considerable amount of training to identify different types of apples. Moreover, stores’ supply is subject to constant changes, which complicates things even further. While these computer vision systems can undeniably reduce cashier headcount, its maintenance costs can add up over time. As with any other technology implementation, it’s critical to assess ROI and system feasibility in a particular business case.

Security And Surveillance

According to the 2019 Retail Crime Survey, retail theft costs the industry more than £700 million yearly. While almost every brick-and-mortar store employs surveillance cameras, the effectiveness of such security systems still largely depends on humans’ attention. With enough training, computer vision software has proven to be far more reliable in detecting shoplifters and recognizing suspicious behavior. Moreover, retail chains can store images of known wrongdoers and instantly recognize them with the help of facial recognition.

One of the most popular theft techniques is tailgating. A very simple computer vision software can eliminate such a threat by comparing the number of employee check-ins with the number of people who have entered the store. An enhanced version of such a system would involve facial recognition software detecting unauthorized persons.

In addition, computer vision software can serve as another layer of security for fire and smoke detection. According to Iflexion, the technology is advanced enough to produce smart video surveillance systems for detecting anomalies in real time and instantly alerting staff if something unusual happens.

Inventory Management

Retail companies can use computer vision to address their inventory management problems, too. This can be done in several different ways. For example, Tally developed by Simbe Robotics is a mobile robot that uses 3D cameras and machine learning to notify staff about out-of-stock products or damaged packaging. Again, this is the case where AI can perform repetitive tasks much faster and with better accuracy than humans.

Trax’s Retail Watch system uses computer-vision-enabled cameras installed on ceilings and shelves to monitor stores’ inventory every hour. Similar to Tally, the ML-powered software then alerts staff about misplaced or missing products. Trax reports that their inventory management system helps retail stores to increase sales by 1%. While this number may not look too impressive for a small family-owned retail shop, it’s a noticeable bump in sales for large retail chains in the long term.

Real-Time Behavioral Analytics

The aforementioned applications of computer vision mostly come down to streamlining existing in-store operations. However, when this technology is coupled with advanced ML algorithms, it can significantly enhance the customer experience.

For example, Swiss computer vision startup Advertima has developed a platform that helps retailers to analyze customers’ experiences in real time and dynamically respond to their needs and wants. The software can accurately identify customers’ age and gender, interpret their emotions and gestures, and analyze what products and store areas grab the most attention at any given time.

This opens up new opportunities for retailers to better understand their customers and provide engaging in-store experiences by delivering targeted in-store advertising. Importantly, Advertima stresses that their technology is GDPR-compliant and doesn’t compromise people’s privacy.

Conclusion

Currently, the most feasible computer vision solutions in retail including self-checkouts, smart surveillance, and automated inventory management systems are poised to augment tasks traditionally performed by humans, improving efficiency, and cutting costs. As the technology moves forward, we are likely to see widespread implementation of Advertima-like systems that will help retailers increase sales and drive advertising ROI up, among other benefits.